vm_fault: improve interface for vm_fault_quick_hold_pages()

[kib]

Summary

Provide (transitional) vm_fault_quick_hold_pages_e() function that
returns distinguishable errors in different failure situation.  Also it
does not panic() on too short array case, but return an error, allowing
sometimes lessen the burden of the check from the caller.

vm_fault_quick_hold_pages() becomes a wrapper, that should be eliminated
eventually.

[markj]

sys/vm/vm_fault.c

2034

EINVAL strikes me as more "correct" there.

2035

Or

count = atop(end - addr);
if (count > max_count)
    return (ENOSPC);

to avoid some minor duplication.

[kib]

sys/vm/vm_fault.c

2035

With count having the int type, I think this would be not correct.

[kib]

s/ENOSPC/EINVAL/

[imp]

Generally I like it, but the name is very long and specific for the logical concept of 'map this into the physical for I/O and keep it there'. But I have no better name, so I'll just grouse because I never could remember the order of the verbs when I was writing the nvme mapping code.

sys/vm/vm_fault.c

2005

s/error is returned/error is returned and any holds are released./ maybe? Either *ALL* the pages are held, or NONE of them are, never some subset (there's code that thinks that a subset can be a possible result in the tree that we should attend to).

2035

Yea, I agree. Logically it makes sense, but in the face of possible overflows, the current conservatism is better.

[markj]

sys/vm/vm_fault.c

2114

This needs to be updated too.

[kib]

Check against EINVAL.
Comment update.

[imp]

thanks for the update.

[alc]

In D52165#1191863, @imp wrote:

Generally I like it, but the name is very long and specific for the logical concept of 'map this into the physical for I/O and keep it there'. But I have no better name, so I'll just grouse because I never could remember the order of the verbs when I was writing the nvme mapping code.

I would suggest dropping "quick_" from the name before anyone actually starts using it. The origin of "quick_" dates back to the Dyson-era, and represents the fact that the code tries to lookup the physical page(s) from the pmap before consulting the machine-independent data structures. Today, I don't see that as being important.